The use of high resolution optical imagery systems, as in photolithographic systems for the semiconductor industry and microscope systems for a wide variety of applications, has continued to grow despite the availability of other technologies, such as high resolution systems based upon atomic particle matter typified by electron beams or X-rays. The greater cost and lower operating convenience of these latter systems, as well as the longer times required for the formation of images, establish that the optical imagery systems will remain preferable for many applications for the foreseeable future. However, constantly increasing requirements for more precise technology have taken the optical imagery systems virtually to the limits of resolution values which can be achieved with refractive optics. For example, very large scale integrated circuits are constantly being reduced in size and made with higher component density, an objective measure of which is the minimum linewidth specification. Whereas one micron linewidths were suitable until recently, present objectives in the industry contemplate linewidths well down into the submicron category, of less than 0.5 micron and even 0.3 micron. This requires a line resolution for a refractive optical system of the order of several thousand lines per millimeter, which has not heretofore been achievable with an optical imaging system of suitable aperture and depth of field.
In response to these problems the optical industry has devised progressively more sophisticated multi-element lens systems using advanced lens design computer programs. The advanced level of the state of the art is exemplified by the so-called "i-line" lens system, which utilizes a complex configuration of some twenty refractive elements of highest quality glasses. The best that this system can achieve, however, is in the range of 0.7 micron linewidth resolution, because the multiple factors involved in complex lens design (chromaticism, coma, astigmatism, spherical aberration being included), together with the problems of achieving sufficient uniformity and adequate wave energy at the target, establish ultimate limitations that are presently at about 0.7 micron linewidth. There are also inherent limitations on manufacture when dealing with this order of precision. For example, the best diamond turning procedures still leave optical surfaces too rough for operation at short wavelengths (e.g. ultraviolet).
The semiconductor industry, however, has devised many production and inspection procedures based upon optical imaging systems, and would prefer to use these for the specific advantages they provide. For example, in preparing the successive layers on a silicon or other wafer, a "wafer stepper" system is employed incorporating the high resolution refractive optics. There is a different precision photomask for each layer to be formed. The wafer is first covered with a layer of photosensitive material of the type on which an image can be fixed by exposure to a suitable amount of light energy. The wafer stepper mechanism then precisely and sequentially places the wafer at chosen matrix positions relative to an optical axis. At each position in the matrix pattern on the wafer, an exposure is made through the photomask, with the optical system typically reducing the image a selected amount, usually five or ten times. Inherent requirements for this type of system are that the light energy be adequate for each exposure, that the exposed image be uniform across the image, that the depth of field be sufficient and that the resolution meet the specifications of the design. These requirements are not easily met in combination, because the very small size of the images and the extreme precision that are required greatly restrict the design alternatives that are available. Once the exposure is made at all positions in the matrix and the unfixed material is washed off, the images can be checked for accuracy and uniformity of reproduction. Optical microscopes are usually employed for checking, on a statistical basis, the characteristics of the various images. The inspection may comprise one or more of a combination of procedures involving automatic or operator measurement of linewidths or other characteristics, but all of these procedures entail precise and high resolution magnification of the image.
The problems of obtaining higher resolution optical imagery systems of practical application thus appear to have approached a limit. Whether or not such limit ultimately is found to be insurmountable with more complex multi-element lens systems remains to be seen. Some substantially different approach appears to be needed, however, that would free optical imagery systems from the constraints on design and manufacture that are inherently imposed in reconciling many higher order terms involved in optical design equations. Tentative movements in this direction were made a number of years ago in proposals that an aspherical element of a special character be introduced into the lens system. These proposals are best evidenced in an article by Kenro Miyamoto entitled "The Phase Fresnel Lens", presented at the November 1960 meeting of the Optical Society of America and subsequently published in the Journal of the Optical Society of America, January 1961, pp. 17-20. In that article, Miyamoto also referenced earlier articles of philosophically similar nature. He principally suggested that a "phase Fresnel lens" be inserted in the pupil plane of an optical system to deform the wavefront surface passing therethrough so as to compensate, for example, for spherical aberration. His proposals were general in nature with no consideration being given to problems of achieving high transmission efficiency, high resolution such as would approach the needs of the semiconductor industry, or adequate depth of field. Miyamoto, in an example, suggested the use of single layer thin film rings of a minimum radial dimension of 0.63 mm. He did not address the difficulties involved in fabricating a more precise system, i.e. a blazed transmission grating.
Miyamoto asserts that a phase Fresnel lens can be made to deform a wave surface by an amount: EQU .phi.(u,v)-(k-1)
where K=1, 2, . . . m, where the amount of deformation in all zones is smaller than .lambda., by depositing a (single) thin film covering various circular zones. He then asserts that a wave surface thus deformed is "quite equivalent" to a lens which deforms the wave surface by the amount .phi.(u,v).
His equations describe a perfectly blazed phase grating yet his description of his method using a single thin film leads to the creation of a binary phase grating, which might also be called a "phase reversal zone plate". This type of grating can only function to provide alternation of phase delays between two values.
The phase reversal zone plate was studied by Melvin H. Horman in an article entitled "Efficiencies of Zone Plates and Phase Zone Plates", published in Applied Optics, November 1967, pp. 2011-2013. Horman defines the efficiency of a zone or phase plate as the "percentage of the flux in the illuminating wavefront which goes to their principal images", and using this definition he gives the first order efficiency of the phase reversal zone plate as 40.5%. Horman notes that the efficiency of a phase Fresnel lens, if it could be built, would approach 100%. Fabrication of a phase Fresnel lens of high efficiency, working in conjunction with highly corrected optics, however, has apparently not been attempted or reported in the intervening years. Triangular profile plates for independent use as microlenses have been made for certain applications.
The Miyamoto proposal thus is recognized as offering the possibility for greater freedom of lens design, but as far as is known from the literature was never implemented. This was probably due to a combination of reasons including limitations perceived as to the advantages to be derived, the difficulty of fabrication of the phase Fresnel lens in the form described, other advances in optical design using solely refractive optics, and a lack of appreciation of much more complex factors inherent in the problem. For example, there can be a substantial variation in efficiency between the parallel and orthogonal components of incident light, with grating blaze angle. Also, Miyamoto failed to appreciate, or at least discuss, the important role that temporal coherence of the individual spectral components plays in maintaining the resolution or space-bandwidth product of a phase Fresnel lens. It is shown hereafter that by properly considering, in the manipulation of wavefront aberration, factors including wave component distributions, the precise distribution of the illuminating energy, and local, temporal and spatial rearrangement of phase relationships, together with a coactive refractive lens configuration, the resolution of an optical imagery or readout system can be increased beyond levels previously thought unattainable, with useful depth of field and high efficiency.
The same principles upon which high resolution optical imagery or readout can be achieved by combinations of phase gratings and optical refractive elements can also be used in other optical applications. These include microscopy and optical transform functions, conical axicon phase gratings in combination with a spherical object lens, cylindrical phase gratings in combination with conventional cylindrical lenses, and toroidal aspheric grating lenses. Conical axicon phase gratings can be particularly useful in combination with optical refractive elements to provide a narrow line of light of predetermined length for an optical disk recording or readout device, eliminating the need for an autofocus system. The ability to precisely correct wavefront aberrations can in other words be of potential benefit wherever refractive optics limits are approached provided that the particular spectral characteristics of phase plates are recognized and accounted for in the system design.